Source file src/encoding/json/stream.go

     1  // Copyright 2010 The Go Authors. All rights reserved.
     2  // Use of this source code is governed by a BSD-style
     3  // license that can be found in the LICENSE file.
     4  
     5  package json
     6  
     7  import (
     8  	"bytes"
     9  	"errors"
    10  	"io"
    11  )
    12  
    13  // A Decoder reads and decodes JSON values from an input stream.
    14  type Decoder struct {
    15  	r       io.Reader
    16  	buf     []byte
    17  	d       decodeState
    18  	scanp   int   // start of unread data in buf
    19  	scanned int64 // amount of data already scanned
    20  	scan    scanner
    21  	err     error
    22  
    23  	tokenState int
    24  	tokenStack []int
    25  }
    26  
    27  // NewDecoder returns a new decoder that reads from r.
    28  //
    29  // The decoder introduces its own buffering and may
    30  // read data from r beyond the JSON values requested.
    31  func NewDecoder(r io.Reader) *Decoder {
    32  	return &Decoder{r: r}
    33  }
    34  
    35  // UseNumber causes the Decoder to unmarshal a number into an interface{} as a
    36  // Number instead of as a float64.
    37  func (dec *Decoder) UseNumber() { dec.d.useNumber = true }
    38  
    39  // DisallowUnknownFields causes the Decoder to return an error when the destination
    40  // is a struct and the input contains object keys which do not match any
    41  // non-ignored, exported fields in the destination.
    42  func (dec *Decoder) DisallowUnknownFields() { dec.d.disallowUnknownFields = true }
    43  
    44  // Decode reads the next JSON-encoded value from its
    45  // input and stores it in the value pointed to by v.
    46  //
    47  // See the documentation for Unmarshal for details about
    48  // the conversion of JSON into a Go value.
    49  func (dec *Decoder) Decode(v any) error {
    50  	if dec.err != nil {
    51  		return dec.err
    52  	}
    53  
    54  	if err := dec.tokenPrepareForDecode(); err != nil {
    55  		return err
    56  	}
    57  
    58  	if !dec.tokenValueAllowed() {
    59  		return &SyntaxError{msg: "not at beginning of value", Offset: dec.InputOffset()}
    60  	}
    61  
    62  	// Read whole value into buffer.
    63  	n, err := dec.readValue()
    64  	if err != nil {
    65  		return err
    66  	}
    67  	dec.d.init(dec.buf[dec.scanp : dec.scanp+n])
    68  	dec.scanp += n
    69  
    70  	// Don't save err from unmarshal into dec.err:
    71  	// the connection is still usable since we read a complete JSON
    72  	// object from it before the error happened.
    73  	err = dec.d.unmarshal(v)
    74  
    75  	// fixup token streaming state
    76  	dec.tokenValueEnd()
    77  
    78  	return err
    79  }
    80  
    81  // Buffered returns a reader of the data remaining in the Decoder's
    82  // buffer. The reader is valid until the next call to Decode.
    83  func (dec *Decoder) Buffered() io.Reader {
    84  	return bytes.NewReader(dec.buf[dec.scanp:])
    85  }
    86  
    87  // readValue reads a JSON value into dec.buf.
    88  // It returns the length of the encoding.
    89  func (dec *Decoder) readValue() (int, error) {
    90  	dec.scan.reset()
    91  
    92  	scanp := dec.scanp
    93  	var err error
    94  Input:
    95  	// help the compiler see that scanp is never negative, so it can remove
    96  	// some bounds checks below.
    97  	for scanp >= 0 {
    98  
    99  		// Look in the buffer for a new value.
   100  		for ; scanp < len(dec.buf); scanp++ {
   101  			c := dec.buf[scanp]
   102  			dec.scan.bytes++
   103  			switch dec.scan.step(&dec.scan, c) {
   104  			case scanEnd:
   105  				// scanEnd is delayed one byte so we decrement
   106  				// the scanner bytes count by 1 to ensure that
   107  				// this value is correct in the next call of Decode.
   108  				dec.scan.bytes--
   109  				break Input
   110  			case scanEndObject, scanEndArray:
   111  				// scanEnd is delayed one byte.
   112  				// We might block trying to get that byte from src,
   113  				// so instead invent a space byte.
   114  				if stateEndValue(&dec.scan, ' ') == scanEnd {
   115  					scanp++
   116  					break Input
   117  				}
   118  			case scanError:
   119  				dec.err = dec.scan.err
   120  				return 0, dec.scan.err
   121  			}
   122  		}
   123  
   124  		// Did the last read have an error?
   125  		// Delayed until now to allow buffer scan.
   126  		if err != nil {
   127  			if err == io.EOF {
   128  				if dec.scan.step(&dec.scan, ' ') == scanEnd {
   129  					break Input
   130  				}
   131  				if nonSpace(dec.buf) {
   132  					err = io.ErrUnexpectedEOF
   133  				}
   134  			}
   135  			dec.err = err
   136  			return 0, err
   137  		}
   138  
   139  		n := scanp - dec.scanp
   140  		err = dec.refill()
   141  		scanp = dec.scanp + n
   142  	}
   143  	return scanp - dec.scanp, nil
   144  }
   145  
   146  func (dec *Decoder) refill() error {
   147  	// Make room to read more into the buffer.
   148  	// First slide down data already consumed.
   149  	if dec.scanp > 0 {
   150  		dec.scanned += int64(dec.scanp)
   151  		n := copy(dec.buf, dec.buf[dec.scanp:])
   152  		dec.buf = dec.buf[:n]
   153  		dec.scanp = 0
   154  	}
   155  
   156  	// Grow buffer if not large enough.
   157  	const minRead = 512
   158  	if cap(dec.buf)-len(dec.buf) < minRead {
   159  		newBuf := make([]byte, len(dec.buf), 2*cap(dec.buf)+minRead)
   160  		copy(newBuf, dec.buf)
   161  		dec.buf = newBuf
   162  	}
   163  
   164  	// Read. Delay error for next iteration (after scan).
   165  	n, err := dec.r.Read(dec.buf[len(dec.buf):cap(dec.buf)])
   166  	dec.buf = dec.buf[0 : len(dec.buf)+n]
   167  
   168  	return err
   169  }
   170  
   171  func nonSpace(b []byte) bool {
   172  	for _, c := range b {
   173  		if !isSpace(c) {
   174  			return true
   175  		}
   176  	}
   177  	return false
   178  }
   179  
   180  // An Encoder writes JSON values to an output stream.
   181  type Encoder struct {
   182  	w          io.Writer
   183  	err        error
   184  	escapeHTML bool
   185  
   186  	indentBuf    *bytes.Buffer
   187  	indentPrefix string
   188  	indentValue  string
   189  }
   190  
   191  // NewEncoder returns a new encoder that writes to w.
   192  func NewEncoder(w io.Writer) *Encoder {
   193  	return &Encoder{w: w, escapeHTML: true}
   194  }
   195  
   196  // Encode writes the JSON encoding of v to the stream,
   197  // followed by a newline character.
   198  //
   199  // See the documentation for Marshal for details about the
   200  // conversion of Go values to JSON.
   201  func (enc *Encoder) Encode(v any) error {
   202  	if enc.err != nil {
   203  		return enc.err
   204  	}
   205  	e := newEncodeState()
   206  	err := e.marshal(v, encOpts{escapeHTML: enc.escapeHTML})
   207  	if err != nil {
   208  		return err
   209  	}
   210  
   211  	// Terminate each value with a newline.
   212  	// This makes the output look a little nicer
   213  	// when debugging, and some kind of space
   214  	// is required if the encoded value was a number,
   215  	// so that the reader knows there aren't more
   216  	// digits coming.
   217  	e.WriteByte('\n')
   218  
   219  	b := e.Bytes()
   220  	if enc.indentPrefix != "" || enc.indentValue != "" {
   221  		if enc.indentBuf == nil {
   222  			enc.indentBuf = new(bytes.Buffer)
   223  		}
   224  		enc.indentBuf.Reset()
   225  		err = Indent(enc.indentBuf, b, enc.indentPrefix, enc.indentValue)
   226  		if err != nil {
   227  			return err
   228  		}
   229  		b = enc.indentBuf.Bytes()
   230  	}
   231  	if _, err = enc.w.Write(b); err != nil {
   232  		enc.err = err
   233  	}
   234  	encodeStatePool.Put(e)
   235  	return err
   236  }
   237  
   238  // SetIndent instructs the encoder to format each subsequent encoded
   239  // value as if indented by the package-level function Indent(dst, src, prefix, indent).
   240  // Calling SetIndent("", "") disables indentation.
   241  func (enc *Encoder) SetIndent(prefix, indent string) {
   242  	enc.indentPrefix = prefix
   243  	enc.indentValue = indent
   244  }
   245  
   246  // SetEscapeHTML specifies whether problematic HTML characters
   247  // should be escaped inside JSON quoted strings.
   248  // The default behavior is to escape &, <, and > to \u0026, \u003c, and \u003e
   249  // to avoid certain safety problems that can arise when embedding JSON in HTML.
   250  //
   251  // In non-HTML settings where the escaping interferes with the readability
   252  // of the output, SetEscapeHTML(false) disables this behavior.
   253  func (enc *Encoder) SetEscapeHTML(on bool) {
   254  	enc.escapeHTML = on
   255  }
   256  
   257  // RawMessage is a raw encoded JSON value.
   258  // It implements Marshaler and Unmarshaler and can
   259  // be used to delay JSON decoding or precompute a JSON encoding.
   260  type RawMessage []byte
   261  
   262  // MarshalJSON returns m as the JSON encoding of m.
   263  func (m RawMessage) MarshalJSON() ([]byte, error) {
   264  	if m == nil {
   265  		return []byte("null"), nil
   266  	}
   267  	return m, nil
   268  }
   269  
   270  // UnmarshalJSON sets *m to a copy of data.
   271  func (m *RawMessage) UnmarshalJSON(data []byte) error {
   272  	if m == nil {
   273  		return errors.New("json.RawMessage: UnmarshalJSON on nil pointer")
   274  	}
   275  	*m = append((*m)[0:0], data...)
   276  	return nil
   277  }
   278  
   279  var _ Marshaler = (*RawMessage)(nil)
   280  var _ Unmarshaler = (*RawMessage)(nil)
   281  
   282  // A Token holds a value of one of these types:
   283  //
   284  //	Delim, for the four JSON delimiters [ ] { }
   285  //	bool, for JSON booleans
   286  //	float64, for JSON numbers
   287  //	Number, for JSON numbers
   288  //	string, for JSON string literals
   289  //	nil, for JSON null
   290  type Token any
   291  
   292  const (
   293  	tokenTopValue = iota
   294  	tokenArrayStart
   295  	tokenArrayValue
   296  	tokenArrayComma
   297  	tokenObjectStart
   298  	tokenObjectKey
   299  	tokenObjectColon
   300  	tokenObjectValue
   301  	tokenObjectComma
   302  )
   303  
   304  // advance tokenstate from a separator state to a value state
   305  func (dec *Decoder) tokenPrepareForDecode() error {
   306  	// Note: Not calling peek before switch, to avoid
   307  	// putting peek into the standard Decode path.
   308  	// peek is only called when using the Token API.
   309  	switch dec.tokenState {
   310  	case tokenArrayComma:
   311  		c, err := dec.peek()
   312  		if err != nil {
   313  			return err
   314  		}
   315  		if c != ',' {
   316  			return &SyntaxError{"expected comma after array element", dec.InputOffset()}
   317  		}
   318  		dec.scanp++
   319  		dec.tokenState = tokenArrayValue
   320  	case tokenObjectColon:
   321  		c, err := dec.peek()
   322  		if err != nil {
   323  			return err
   324  		}
   325  		if c != ':' {
   326  			return &SyntaxError{"expected colon after object key", dec.InputOffset()}
   327  		}
   328  		dec.scanp++
   329  		dec.tokenState = tokenObjectValue
   330  	}
   331  	return nil
   332  }
   333  
   334  func (dec *Decoder) tokenValueAllowed() bool {
   335  	switch dec.tokenState {
   336  	case tokenTopValue, tokenArrayStart, tokenArrayValue, tokenObjectValue:
   337  		return true
   338  	}
   339  	return false
   340  }
   341  
   342  func (dec *Decoder) tokenValueEnd() {
   343  	switch dec.tokenState {
   344  	case tokenArrayStart, tokenArrayValue:
   345  		dec.tokenState = tokenArrayComma
   346  	case tokenObjectValue:
   347  		dec.tokenState = tokenObjectComma
   348  	}
   349  }
   350  
   351  // A Delim is a JSON array or object delimiter, one of [ ] { or }.
   352  type Delim rune
   353  
   354  func (d Delim) String() string {
   355  	return string(d)
   356  }
   357  
   358  // Token returns the next JSON token in the input stream.
   359  // At the end of the input stream, Token returns nil, io.EOF.
   360  //
   361  // Token guarantees that the delimiters [ ] { } it returns are
   362  // properly nested and matched: if Token encounters an unexpected
   363  // delimiter in the input, it will return an error.
   364  //
   365  // The input stream consists of basic JSON values—bool, string,
   366  // number, and null—along with delimiters [ ] { } of type Delim
   367  // to mark the start and end of arrays and objects.
   368  // Commas and colons are elided.
   369  func (dec *Decoder) Token() (Token, error) {
   370  	for {
   371  		c, err := dec.peek()
   372  		if err != nil {
   373  			return nil, err
   374  		}
   375  		switch c {
   376  		case '[':
   377  			if !dec.tokenValueAllowed() {
   378  				return dec.tokenError(c)
   379  			}
   380  			dec.scanp++
   381  			dec.tokenStack = append(dec.tokenStack, dec.tokenState)
   382  			dec.tokenState = tokenArrayStart
   383  			return Delim('['), nil
   384  
   385  		case ']':
   386  			if dec.tokenState != tokenArrayStart && dec.tokenState != tokenArrayComma {
   387  				return dec.tokenError(c)
   388  			}
   389  			dec.scanp++
   390  			dec.tokenState = dec.tokenStack[len(dec.tokenStack)-1]
   391  			dec.tokenStack = dec.tokenStack[:len(dec.tokenStack)-1]
   392  			dec.tokenValueEnd()
   393  			return Delim(']'), nil
   394  
   395  		case '{':
   396  			if !dec.tokenValueAllowed() {
   397  				return dec.tokenError(c)
   398  			}
   399  			dec.scanp++
   400  			dec.tokenStack = append(dec.tokenStack, dec.tokenState)
   401  			dec.tokenState = tokenObjectStart
   402  			return Delim('{'), nil
   403  
   404  		case '}':
   405  			if dec.tokenState != tokenObjectStart && dec.tokenState != tokenObjectComma {
   406  				return dec.tokenError(c)
   407  			}
   408  			dec.scanp++
   409  			dec.tokenState = dec.tokenStack[len(dec.tokenStack)-1]
   410  			dec.tokenStack = dec.tokenStack[:len(dec.tokenStack)-1]
   411  			dec.tokenValueEnd()
   412  			return Delim('}'), nil
   413  
   414  		case ':':
   415  			if dec.tokenState != tokenObjectColon {
   416  				return dec.tokenError(c)
   417  			}
   418  			dec.scanp++
   419  			dec.tokenState = tokenObjectValue
   420  			continue
   421  
   422  		case ',':
   423  			if dec.tokenState == tokenArrayComma {
   424  				dec.scanp++
   425  				dec.tokenState = tokenArrayValue
   426  				continue
   427  			}
   428  			if dec.tokenState == tokenObjectComma {
   429  				dec.scanp++
   430  				dec.tokenState = tokenObjectKey
   431  				continue
   432  			}
   433  			return dec.tokenError(c)
   434  
   435  		case '"':
   436  			if dec.tokenState == tokenObjectStart || dec.tokenState == tokenObjectKey {
   437  				var x string
   438  				old := dec.tokenState
   439  				dec.tokenState = tokenTopValue
   440  				err := dec.Decode(&x)
   441  				dec.tokenState = old
   442  				if err != nil {
   443  					return nil, err
   444  				}
   445  				dec.tokenState = tokenObjectColon
   446  				return x, nil
   447  			}
   448  			fallthrough
   449  
   450  		default:
   451  			if !dec.tokenValueAllowed() {
   452  				return dec.tokenError(c)
   453  			}
   454  			var x any
   455  			if err := dec.Decode(&x); err != nil {
   456  				return nil, err
   457  			}
   458  			return x, nil
   459  		}
   460  	}
   461  }
   462  
   463  func (dec *Decoder) tokenError(c byte) (Token, error) {
   464  	var context string
   465  	switch dec.tokenState {
   466  	case tokenTopValue:
   467  		context = " looking for beginning of value"
   468  	case tokenArrayStart, tokenArrayValue, tokenObjectValue:
   469  		context = " looking for beginning of value"
   470  	case tokenArrayComma:
   471  		context = " after array element"
   472  	case tokenObjectKey:
   473  		context = " looking for beginning of object key string"
   474  	case tokenObjectColon:
   475  		context = " after object key"
   476  	case tokenObjectComma:
   477  		context = " after object key:value pair"
   478  	}
   479  	return nil, &SyntaxError{"invalid character " + quoteChar(c) + context, dec.InputOffset()}
   480  }
   481  
   482  // More reports whether there is another element in the
   483  // current array or object being parsed.
   484  func (dec *Decoder) More() bool {
   485  	c, err := dec.peek()
   486  	return err == nil && c != ']' && c != '}'
   487  }
   488  
   489  func (dec *Decoder) peek() (byte, error) {
   490  	var err error
   491  	for {
   492  		for i := dec.scanp; i < len(dec.buf); i++ {
   493  			c := dec.buf[i]
   494  			if isSpace(c) {
   495  				continue
   496  			}
   497  			dec.scanp = i
   498  			return c, nil
   499  		}
   500  		// buffer has been scanned, now report any error
   501  		if err != nil {
   502  			return 0, err
   503  		}
   504  		err = dec.refill()
   505  	}
   506  }
   507  
   508  // InputOffset returns the input stream byte offset of the current decoder position.
   509  // The offset gives the location of the end of the most recently returned token
   510  // and the beginning of the next token.
   511  func (dec *Decoder) InputOffset() int64 {
   512  	return dec.scanned + int64(dec.scanp)
   513  }
   514  

View as plain text